Research map / Learning theory and algorithms / Prediction and decision losses
The Nyström method for convex loss functions
Andrea Della Vecchia, Ernesto De Vito, Jaouad Mourtada, Lorenzo Rosasco
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace linear SVM with Nyström-Pegasos | SVM · SUSY | classification error: 28.1% -> 20.0% | Improved |
| replace linear SVM with Nyström-Pegasos | SVM · Mnist | classification error: 12.4% -> 2.2% | Improved |
| replace linear SVM with Nyström-Pegasos | SVM · CIFAR | classification error: 31.5% -> 19.2% | Improved |
| use Nyström-Pegasos instead of exact kernel SVM | K-SVM · Mnist | training time (s): 1601 -> 1342 | Improved |
| use Nyström-Pegasos instead of exact kernel SVM | K-SVM · Webspam | training time (s): 6044 -> 2440 | Improved |
| use Nyström-Pegasos instead of exact kernel SVM | K-SVM · CIFAR | training time (s): 6339 -> 2408 | Improved |
| use Nyström-Pegasos instead of exact kernel SVM | K-SVM · a9a | training time (s): 114 -> 29.3 | Improved |
| use Nyström-Pegasos instead of exact kernel SVM | K-SVM · Usps | training time (s): 4.4 -> 19.8 | Hurt |
| use Nyström-Pegasos instead of exact kernel SVM | K-SVM · Mnist | classification error: 2.2% -> 2.2% | No clear effect |
| use Nyström-Pegasos instead of exact kernel SVM | K-SVM · Webspam | classification error: 1.1% -> 1.3% | No clear effect |
| use hinge loss (Nyström-Pegasos) instead of square loss (Nyström-KRR) | Nyström ERM · Mnist | classification error: 2.4% -> 2.2% | Improved |
| use hinge loss (Nyström-Pegasos) instead of square loss (Nyström-KRR) | Nyström ERM · a9a | classification error: 14.9% -> 15.1% | Hurt |
Closest papers by meaning
- Sampling-based Nyström Approximation and Kernel Quadrature (ICML 2023 · similarity 0.89)
- Nyström Kernel Mean Embeddings (ICML 2022 · similarity 0.88)
- Revisiting the Nyström Method for Improved Large-scale Machine Learning (JMLR 2016 · similarity 0.88)
- The Geometry and Calculus of Losses (JMLR 2023 · similarity 0.88)
- Optimal Convergence Rates for Agnostic Nyström Kernel Learning (ICML 2023 · similarity 0.87)
- Robust high dimensional learning for Lipschitz and convex losses (JMLR 2020 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
